Constructing the Hyper-Kamiokande Computing Model in the Build Up to Data Taking

TL;DR

This paper details the development of a scalable distributed computing model for Hyper-Kamiokande, a next-generation neutrino experiment, leveraging existing grid infrastructure to handle massive data and simulation demands forecasted for 2036.

Contribution

It introduces a comprehensive computing framework utilizing Worldwide LHC Grid and GridPP DIRAC to meet Hyper-Kamiokande's large-scale data processing needs.

Findings

Estimated data volume of 35PB per replica by 2036

Forecasted requirement of 8,700 CPU cores (~100,000 HS06)

Implementation of workflow and resource management tools

Abstract

Hyper-Kamiokande is a next-generation multi-purpose neutrino experiment with a primary focus on constraining CP-violation in the lepton sector. It features a diverse science programme that includes neutrino oscillation studies, astrophysics, neutrino cross-section measurements, and searches for physics beyond the standard model, such as proton decay. Building on its predecessor, Super-Kamiokande, the Hyper-Kamiokande far detector has a total volume approximately 5 times larger and is estimated to collect nearly 2PB of data per year. The experiment will also include both on- and off-axis near detectors, including an Intermediate Water Cherenkov Detector. To manage the significant demands relating to the data from these detectors, and the associated Monte Carlo simulations for a range of physics studies, an efficient and scalable distributed computing model is essential. This model leverages Worldwide LHC Grid computing infrastructure and utilises the GridPP DIRAC instance for both workload management and for file cataloguing. In this report we forecast the computing requirements for the Hyper-K experiment, estimated to reach around 35PB (per replica) and 8,700 CPU cores (~ 100,000 HS06) by 2036. We outline the resources, tools, and workflow in place to satisfy this demand.